Methods | Special Issue: Network Psychometrics

Comparing maximum likelihood and maximum pseudolikelihood estimators for the Ising model

Sara Keetelaar

, Nikola Sekulovski

, Denny Borsboom

, & Maarten Marsman

https://doi.org/10.56296/aip00013
Published: April 20, 2024
Copyright: The authors (CC BY 4.0)

Keetelaar, S., Sekulovski, N., Borsboom, D., & Marsman, M. (2024). Comparing maximum likelihood and maximum pseudolikelihood estimators for the Ising model. advances.in/psychology, 2, e25745. https://doi.org/10.56296/aip00013

Keetelaar, Sara, et al. "Comparing maximum likelihood and maximum pseudolikelihood estimators for the Ising model." advances.in/psychology, vol. 2, no. 1, 2024, e25745. https://doi.org/10.56296/aip00013.

Keetelaar, Sara, Nikola Sekulovski, Denny Borsboom, and Maarten Marsman. 2024. "Comparing maximum likelihood and maximum pseudolikelihood estimators for the Ising model." advances.in/psychology 2 (1): e25745. https://doi.org/10.56296/aip00013.

Keetelaar S, Sekulovski N, Borsboom D, Marsman M. Comparing maximum likelihood and maximum pseudolikelihood estimators for the Ising model. advances.in/psychology. 2024;2(1):e25745. doi:10.56296/aip00013.

Keetelaar, S. et al. (2024) 'Comparing maximum likelihood and maximum pseudolikelihood estimators for the Ising model', advances.in/psychology, 2(1), e25745. Available at: https://doi.org/10.56296/aip00013.

Download .RIS Download .bib

The Ising model is one of the most popular models in network psychometrics. However, statistical analysis of the Ising model is difficult due to the presence of its intractable normalizing constant in the probability function. As a result, maximum likelihood estimation using the exact likelihood is only possible for small graphs, and approximation methods are needed for larger graphs. Two popular approximations of the exact likelihood are the joint pseudolikelihood (JPL) and the disjoint pseudolikelihood (DPL). These approximations yield consistent estimators, but we do not know how well they perform for finite data. In this paper, we investigate the relative performance of parameter estimation methods based on the two approximations and compare them to maximum likelihood estimation using the exact likelihood. We perform an extensive simulation study comparing the estimators in terms of bias and variance. We show that maximum pseudolikelihood estimation based on the JPL is a stable estimation method that is able to accurately approximate the maximum likelihood estimates, but that maximum pseudolikelihood estimation based on the DPL only works well for large sample sizes.

Francis K. C. Hui, Ding Ding (2026). A clustered and sparse ising regression model for multivariate binary data. Statistics and Computing. https://doi.org/10.1007/s11222-026-10888-8
Nikola Sekulovski, Tessa F. Blanken, Jonas M B Haslbeck, Maarten Marsman (2025). The impact of dichotomization on network recovery. Behavior Research Methods. https://doi.org/10.3758/s13428-025-02861-6
Ludovica De Carolis, Minjeong Jeon (2025). Network Approaches to Binary Assessment Data: Network Psychometrics Versus Latent Space Item Response Models. Educational and Psychological Measurement. https://doi.org/10.1177/00131644251371187
Elena S. Joukova, S.L. Artemenkov, Bartolome Paul Vincent A (2025). Assessment of students’ subject knowledge based on the results of the Olympiad “Nature of Russia”. Bulletin of Psychological Practice in Education. https://doi.org/10.17759/bppe.2025220210
Karoline Huth, W. Bradford DeLong, Lourens Waldorp, Maarten Marsman, Mijke Rhemtulla (2025). Nodewise Parameter Aggregation for Psychometric Networks. Multivariate Behavioral Research. https://doi.org/10.1080/00273171.2025.2450648
Maarten Marsman, Don van den Bergh, Jonas M B Haslbeck (2025). Bayesian Analysis of the Ordinal Markov Random Field. Psychometrika. https://doi.org/10.1017/psy.2024.4
Giovanni Briganti, Marco Scutari, Sacha Epskamp, Denny Borsboom, Ria H. A. Hoekstra, Hudson Golino, Alexander P. Christensen, Yannick Morvan, Omid V. Ebrahimi, Giulio Costantini, Alexandre Heeren, Jill de Ron, Laura F. Bringmann, Karoline Huth, Jonas M B Haslbeck, Adela‐Maria Isvoranu, Maarten Marsman, Tessa F. Blanken, Allison Gilbert, Teague R. Henry, Eiko I. Fried, Richard J. McNally (2024). Network analysis: An overview for mental health research. International Journal of Methods in Psychiatric Research. https://doi.org/10.1002/mpr.2034
Nikola Sekulovski, Sara Keetelaar, Jonas M B Haslbeck, Maarten Marsman (2023). Sensitivity Analysis of Prior Distributions in Bayesian Graphical Modeling: Guiding Informed Prior Choices for Conditional Independence Testing. https://doi.org/10.31234/osf.io/6m7ca

Download PDF Back to article